Files
Abstract
Timber inventories are designed to give a description of the forest at a 1) determined spatial scale, for a 2) specific area, 3) with a certain level of (un)certainty in mind. Furthermore, while their intended uses may differ, the underlying reason all inventories are made is to collect information. How that information is used, on the other hand, is often much more complicated. It is a statistical reality that data from different sources with differing statistical and sampling characteristics cannot be pooled together for the purpose of deriving a new unbiased estimator. This means, although there is abundant information about our environment (i.e. atmospheric conditions, soil composition, nearly 40 years of satellite imagery, and a wealth of site specific studies sampling for various data) one cannot assimilate these data and produce a new, unbiased estimate for the variable and area of interest. To address this issue, I present three case studies relating to the use of seemingly unrelated and incompatible data for the derivation and application of a high-resolution inventory for the detailed analysis of fiber supply and policy analysis within spatially explicit, stand level areas. In the first study, I fill the data gaps present in the SLC-off Landsat 7 Enhanced Thematic Mapper Plus satellite imagery using the nearest neighbor methods applied to multi-temporal Landsat 5 Thematic Mapper data. The second is an application of modeling forest variables across a series of Landsat imagery for the small-area assessment of streamside management zones and road beautifying buffers. The third study describes the development of a high-resolution forest inventory for the state of Georgia from data that is traditionally not used jointly for predictions. The final inventory retains the statistical integrity of the large-area USDA Forest Inventory and Analysis while maintaining the local accuracy of the small-area timber inventories from our industry partners.